May 27, 2023, 5:13 PM ETInterview, Fireside Chat

Tomer Cohen: Why LinkedIn Stories Failed; How LinkedIn's Feed Was Born; AI Startups | E1019

Current AI models are primarily focused on reconstructing and predicting based on existing public knowledge (the internet).
The next frontier involves AI hypothesizing and generating new knowledge, such as solving mysteries like dark matter, dark energy, or Alzheimer's disease.
This shift represents a transition from models that restructure existing information to those capable of creating entirely new scientific discoveries.
Tomer Cohen views this transition as a fundamental change for the economy, business, and society, moving beyond prediction to innovation.

Background: Cohen's career began in semiconductors (chip design), moved to embedded software, and included founding a consumer company before joining LinkedIn.
Joining LinkedIn: He met Reid Hoffman at a 2008 Stanford lecture on social networks; Hoffman's vision of professional communities as an economic growth engine deeply resonated with him.
Appointment: Cohen joined LinkedIn in 2012 to rebuild the product for mobile and was appointed CPO in 2020.
Mentorship: He cites Reid Hoffman as a mentor who helps him "think in orange" by offering philosophical insights and third-party perspectives on complex problems.
Product Philosophy: Cohen rejects the dichotomy of product as art vs. science, arguing they are interwoven; science provides the craft and data, while art provides vision, intuition, and creativity.
Role Typology: He identifies with all three product profiles (Visionary, Craftsman, Operator), emphasizing the joy of both high-level strategy and the "grind" of execution and iteration.

Job-to-be-Done (JTBD): LinkedIn shifted from segmenting by audience to understanding the emotional and social "jobs" of members, such as the stress of B2B buying requiring consensus or the desire for economic opportunity through creation.
Feed Redesign: A controversial 2015 decision pivoted the feed from an organizational chart/activity feed to a place for professional conversations, initially cannibalizing internal discovery metrics to improve user engagement.
Quality over Growth: During a period of record virality, the team made the difficult decision to curb growth to eliminate spam and shallow engagement, prioritizing trust and quality.
Network vs. Follow: LinkedIn introduced a "follow" relationship (unlike "connect") to handle the 200% year-over-year growth in following non-contacts, clarifying that "connect" is for bidirectional professional help while "follow" is for learning.
Stories Experiment: The launch of "Stories" (ephemeral content) was a hypothesis test that failed because the JTBD for creators was actually permanence and reputation building, not ephemerality.
Founders and Innovation: Cohen argues that while founders are critical for the "first founding moment" of a startup, their role in mature companies (like Microsoft or LinkedIn) shifts to guiding the "spirit" and vision rather than operating the product daily.

Product Leadership: Cohen asserts that AI is now the "engine" of product; leaders must personally learn AI skills (prompting, fine-tuning) to navigate their teams, rather than delegating it to a horizontal team.
Bundling Trend: He predicts a future wave of bundled AI products where a single "master brain" automates complex workflows across multiple traditional software interfaces, reversing the previous trend of unbundling.
Value Accrual: While proprietary data remains a differentiator for startups, the baseline advantage has shifted to the ability to fine-tune foundational models with unique data to create specialized applications.
Skill Sets: The most critical new skill is "Prompt Engineering" or the ability to communicate effectively with AI to guide its output.
Incumbents vs. Startups: Startups have an advantage in re-imagining problems from scratch, but incumbents leverage existing data and infrastructure; the future will likely see incumbents innovating rapidly (e.g., Microsoft's OpenAI integration).
Future of Work: Education must prioritize a "growth mindset" and the ability to learn over specific technical skills, as the half-life of skills is accelerating (estimated at 25% change every 5 years).

Product Jams: Cohen renamed "Product Reviews" to "Product Jams" to shift the culture from seeking validation to providing feedback, held in-person to maximize creative energy.
Briefback Process: Teams summarize session feedback and action items with ETAs to ensure accountability and clarity of execution.
Accountability Model: Decisions utilize a framework distinguishing the Recommender (expertise), Approver (objective alignment), and Decider (final authority) to speed up escalations.
Clarity over Luck: The core philosophy is being "wrong but not confused," meaning a leader should have high conviction and clarity of thought even if the specific hypothesis fails, to ensure learnings are actionable.
Revenue vs. Innovation: LinkedIn balances short-term revenue targets with long-term innovative bets through a continuous planning process that aligns all activities with the vision of "economic opportunity for every member."

Favorite Hiring Question: Candidates are asked about their most complex problem (assessing clarity and depth) and their biggest failure (assessing growth mindset and learning).
Most Embarrassing Release: Cohen cites "Instant Articles" (early mobile caching) as a concept that was ahead of its time and suffered from execution nuances, though he views the failure as a learning opportunity.
Best Advice for New PMs: To move from linear to exponential success, PMs must understand the holistic LinkedIn ecosystem rather than just their specific "swim lane."
Recent Impressive Strategy: He cites Microsoft's integration of OpenAI into its ecosystem as a remarkable example of strategic execution and responsible AI deployment.